ECCM15 - 15TH EUROPEAN CONFERENCE ON COMPOSITE MATERIALS, Venice, Italy, 24-28 June 2012

Study of delamination of five harness satin weave composites molded by hot

pressing. The example of high strength carbon fiber and PEEK matrix.

G. Kamgaing*, P. Nimdum, J. Renard

Centre des Matériaux P. M. Fourt, Mines-Paris Tech, CNRS UMR 7633, BP 87, F-91003 Evry cedex
*georges.kamgaing@mines-paristech.fr

Keywords: five harness satin weave, delamination criterion identifcation, acoustic emission,
mesoscopic scale modeling.

Abstract
The purpose of this study is to identify a delamination criterion and find a way to apply it
directly on industrial parts. Criterion parameters are identified using edge delamination tests
and observations. The experimental setup is equipped with a camera and a data recording
system of acoustic emissions to detect the onset of the delamination and its propagation. The
identified criterion is then validated on a laboratory sample and applied to a real part
modeled at mesoscopic scale (ply scale). These results give interesting prospects for a less
conservative design. In fact, since the appearance of a macroscopic delamination coincides
with the start of a fall of mechanical properties, a good prediction of this phenomenon would
significantly reduce safety factors applied to aerospace components today.

1. Introduction
The prediction of delamination is a complex problem that has interested many researchers.
Early work on this subject started in 80s with Herakovich [1] which showed that this defect
occurs preferentially at the + θ /-θ interfaces of laminates subjected to uniaxial tension. Then
Kim and Soni [2] were proposed the formulation of a quadratic and non-local delamination
criterion. This criterion has then been modified by several authors [3]. Brewer and Lagace [3]
demonstrated the sensitivity of the delamination onset and propagation with increasing
laminate thickness. Based on the work of Brewer and co-workers [4] have proposed a method
based on edge delamination test (EDT) to identify the criteria parameters. More recently,
Nimdum [5] and Bassery [6] modified the Kim’s criterion by taking into account the
influence of friction between laminate plies on delamination strength. Also in the 90s, another
approach for the prediction of delamination based on the fracture mechanics has been
developed a lot in composites. This approach considers the delamination as a crack and
measures the energy release rate during its propagation. Bathias [7] has reviewed full of these
methods. In comparison with Kim’s approach, it fairly requires complex tests, so called DCB
tests, to result in more expensive to perform.
Despite these numerous studies, delamination remains a crucial question. In fact, the majority
of the conventional design criteria of composite components are based on the final failure of
the laminates. This approach necessarily requires a high safety factor which finally is an
ignorance factor. As well-known, before the final fracture, several damage mechanisms occur
in the material. It would be judicious to act earlier, for example at the delamination onset
which is the prior phase to the fracture. This approach has been implemented successfully by
1
ECCM15 - 15TH EUROPEAN CONFERENCE ON COMPOSITE MATERIALS, Venice, Italy, 24-28 June 2012

Nimdum [5] and Bassery [6]. However, these approaches were validated only on laboratory
test specimens. In this reason, the main aims of this investigate are, firstly, to identify the
delamination criterion for the five harness carbon/PEEK satin weave, and then to propose a
generic method which can be applied for industrial structures..

2. Materials
The material of this study is carried out on high strength carbon HTA/PEEK composite
material. The weave is a five harness satin balanced in warp and weft directions and the
thermoplastic PEEK matrix is Evonik’s type. Laminates are manufactured by Carbone Forge
Company using a hot pressing process of impregnated fabrics, so-called prepregs. Prepregs
are 3K type with an area density of 477g/m2 and a fiber weight fraction of 0.6. After molding,
the laminates have an average ply thickness of 0.3mm and 0.52 as fiber volume fraction.
To evaluate the orthotropic stiffness matrix of a single ply, necessary for finite element
calculations, only six elastic constants E1, E3, G12, G13, ν12 and ν13 (where the subscripts 1, 2
and 3 indicate respectively warp, weft and thickness directions) are required since warp and
weft directions are equivalent. E1, G12, G13 and ν12 are experimentally obtained by tensile,
bending and intralaminar shear tests on 2mm thick laminates. To determine the Young
Modulus E3 in thickness direction, a compression test on 4 mm thick cylindrical rod was
carried out with the assumption that the Young modulus in tensile and compression are
equivalent. Due to the high difficulty to measure the ν13 by experimental test, we choose this
value to be 0.4 which commonly used for this type of material. Finally, the single ply
properties are summarized in table 1.

E1 [MPa] E3 [MPa] G12[MPa] G13[MPa] ν12 ν13

57600 8350 4500 1510 0,05 0,40

Table 1. Summary of the elastic properties of the elementary ply

3. Criterion and Identification.

3.1. Criterion.
In the same manner as cracks growth, delamination occurs in 3 modes including one opening
mode (mode 1) and two shear modes (modes 2 and 3). To predict its initiation, we used an
approach based on the stress tensor component. In comparison with the fracture mechanics
approach, this one is easier to use by designers since the stress field are directly available over
the finite element calculations. Kim and Soni in 1984 [2] were, firstly, to propose a
formulation of the criterion given in equation (1):

σ xz 2 σ yz 2 1 1 σ zz2
( ) + ( ) + σ ( − ) + ( ) =1 (1)
σ xzD σ yzD σ zzDT σ zzDC σ zzDT σ zzDC
ZZ

Where σ xzD and σ yzD represent the delamination strenght in shear modes (modes 2 and 3),
σ zzDT the delamination strenght in opening mode (mode 1 in tension),
σ zzDC the delamination strenght in compression mode (mode 1 in compression),
y
1 0
σ ij = σ ij dy the average stress of σ ij at y0 distance from the free edge.
y 0 ∫0
As can be seen, this criterion is an interactive one as Tsai criteria but introduced the concept
of average stress over a critical distance, making it a non-local criterion. Beside, it also takes
in account the compression effect to tension in open mode. This criterion will be simplified by
2
ECCM15 - 15TH EUROPEAN CONFERENCE ON COMPOSITE MATERIALS, Venice, Italy, 24-28 June 2012

omitting the compression term in mode I. By taking into account the Coulomb effects, the
delamination strength in shear mode is affected by normal compression force which generates
frictions between layers. Then, we are able to simplify this criterion given by equation (2)
where k1 and k2 are friction coefficients in shear modes 2 and 3 respectively.

σ xz 2 σ yz 2 σ zz σ xz σ yz σ zz
( ) + ( ) + = 1 if σ > 0 or ( ) 2
+ ( ) 2 + DC = 1 if σ zz < 0 (2)
σ xz σ yz σ zz σ xz − k1σ zz σ yz − k 2σ zz σ zz
D D DT zz D D

By equation (2), we can make one more assumption to further simplify. First we assume that
all modes of delamination growth have the same critical distance y0. Then, we assume that the
two shear modes are equivalent, i.e. σ xzD = σ yzD and k1 = k2 = k. Finally, Only 4 parameters
will be identified, namely σ zzDT , σ xzD , y0 and k.

3.2. Parameters identification

3.2.1. σ zzDT parameter.
In order to measure this parameter, we need to apply a load at the interface in opening mode.
To do that, the Arcan-Mines test as describe in ref. [6] was realized. However, this test is
extremely difficult to implement. Then, we propose a new system which is more simple as
described in figure 1. Cylindrical samples of 4 mm in thickness and 22 mm in diameter are
joined carefully on an aluminum support with special glue for thermoplastic polymers. In
order to endure the failure at interface layer, not only a notched specimen with 2mm in
thickness and 5mm in width was cut around the rod but also a special system for engaging the
lips of the notched sample is added. Then, this specimen was performed by Instron machine
in thickness direction under tensile test.

Figure 1. Delamination test specimen in opening mode (right) and results (left)

Figure 1 shows the important role of the hanging system (red curves) resulted in high failure
strength in comparison with not using this system. Finally, these results allow us to identify
the delamination strength in open mode to 26 MPa (average value).

3.2.2. y0, σ yzD and k parameters

To assess theses parameters, we used the method proposed by Lecuyer and co-workers which
have been widely used in numerous studies [4,5,8]. Fist, edge delamtination tests (EDTs)
were performed on specific laminates which has a tendency to present delamination in shear
mode. The delamination onset stress was recorded. Then, by using numerical results which
simulate at similar macroscopic stress level in experimental result, the delamination strength
was determined.

3
ECCM15 - 15TH EUROPEAN CONFERENCE ON COMPOSITE MATERIALS, Venice, Italy, 24-28 June 2012

3.2.2.1. Experimental measurement of the delamination load.

As explained in above, we selected (0°,±20°n)s and (0°,±30°n)s laminates with n = 1 and 2,
which are known to be very sensitive to shear delamination at +θ/-θ interfaces. The tensile
specimens are prepared according to European standard EN2561 and the free edge was
polished in order to easier observe by a camera during testing. The test samples are also
equipped with acoustic emission system for monitoring various damage which occur during
load applied. A crosshead travel rates of 0.25mm/min was chosen in order to prevent
saturation of the acoustic recorder system. More information about the acoustic emission test
is available in [9].

Figure 2. In situ observation of the (0°,±20°)s laminate and Stress-strain curves for 2mm/min load rate. In black
in the table, values for 2mm/min; in red values for 0.25 mm/min; in blue values from laminate theory

Figure 2 shows the mechanical properties of laminates at different load rate. These results
which are close to those obtained by applying laminate theory validate single ply properties
(see in the table 1) that we used in finite element simulation.
Figure 2 also shows camera in situ observations of the (0°,±20°)s laminate and illustrates the
material degradation process until failure. The picture Figure 2(a) represents the material prior
to apply load. We can notice that interfaces between adjacent layers are not regular plans as in
unidirectional laminates. Up to about 40% of failure stress, the material microstructure
remains unchanged until 40% (picture (b)) first crack growth occurs on weft (transverse)
yarns of 0° layers. This is quite logical fact since these layers carry most of the load lead to be
first affect. Beside, we also found the discrete white marks occurring at the interfaces that we
assume to be fibre/matrix debonding (micro-delaminations). When the loading continues
increase, these white marks become more and more numerous and seem to coalesce to form
nearly continuous lines. Up to around the 60% of failure stress (Figure2(c)), the first crack
occurs on yarns of 20° layer. This phenomenon represents to the limitation of load transfer to
the 0° layers. When the loading continues increase up to about 95% of failure stress (Figure
2(e)), 0° and 20° layers show several open transverse cracks in their weft yarns. Figure 2(f)
shows the specimen just before failure in an area near the breaking zone. We found that the
significant delamination appear in this zone to confirm the important role of the delamination.
Figure 2(d) show the optical microscopic observation after the specimen failure to confirm the
4
ECCM15 - 15TH EUROPEAN CONFERENCE ON COMPOSITE MATERIALS, Venice, Italy, 24-28 June 2012

white marks which were observed by camera. In the similar procedure as (0°,±20°)s stacking
sequence, five layers of (0°,±30°) laminate were done.

(a) (b)

(c) (d)

Figure 3. Spectra of number of high energy and long duration events versus stress applied for. (a) (0°,±20°)s;
(b) (0°,±20°2)s; (c) (0°,±30°)s; (d) (0°,±30°2)s. Arrows indicate the beginning of propagation.

According to these results, we can deduce that propagation of the delamination of five layers
laminates occurs around 90% of their axial strength (values in red line in figure 4). In fact, we
think that any propagation can’t occur before transverse cracks saturation in the material and
the appearance of several open cracks in 20° or 30° layers which finally act to delay the
propagation of delamination to correspond to the summary of Bathias’s work [7]. For 8 layers
laminates (0°,±20°2)s and (0°,±30°2)s), this saturation stage occurs very early to around 60%
of their longitudinal tensile strength which confirms the thickness effects as shown in Brewer
and Lagace [3]. This summary should be verified by acoustic emission monitored.
By using data recording of acoustic emissions (number, duration, amplitude, frequency,
energy and cumulative energy of events), the duration and energy are good parameters to
represent the damage developed during the tensile loading. We found that delamination
produce the long duration and high energy signal since it involves high energy release to
create the separations of large surfaces. Figure 3 shows the cumulative number events of high
energy signals as function of applied stress. The step increasing level in first stage represents
an isolated delamination which occurs at intersection of warp yarn. When the cracks density
will be enough in the material, these isolated delaminations will coalesce and the macroscopic
propagation will start. This phenomenon will correspond to sudden increase in event numbers.
Finally, we can determine the delamination onset stress around 740 MPa, 490 MPa, 620 MPa
and 360 MPa for (0°,±20°)s, (0°,±30°)s, (0°,±20°2)s and (0°,±30°2)s laminates, respectively.
These values are good agreement with previously results by camera observations.

3.2.2.2. Simulations of tensile tests and parameter’s identification.

This section show the simulation results of the tensile tests performed on the laminates at
delamination onset stress (previously identified). Figure 4 shows the mesh by using 8 Gauss
points of linear cubic elements (C3D8). On a small area close to the free edge, the mesh
refinement is done with element’s size equal to one sixth of the thickness. Then two types of
nodes lines are created: one at free edge in the thickness direction (z direction). The second
one at +θ/–θ interface in y direction along which will be applied the criterion given by
5
ECCM15 - 15TH EUROPEAN CONFERENCE ON COMPOSITE MATERIALS, Venice, Italy, 24-28 June 2012

equation (2). At each node of the lines are stored coordinates and local stresses. Another script
retrieves these data and automatically computes the average stress along the nodes lines.
Finite element simulations are performed with commercial software Zebulon developed at
“Centre Des Matériaux de l’école des mines Paris”. Results are shown in Figure 5.

Figure 4. 3D Mesh of tests specimens for finite element calculations

The first picture (a) shows the friction case which exist the normal compression stresses at all
interfaces. According to Lecuyer’s work [4] proposed that the critical distance and the shear
mode delamination strength are materials properties, then all curves should intersect at the
same point where this coordinates should correspond to these two parameters. However, as
can be seen the second picture (b), all curves intersect at the same abscissa which is about
0.3mm but at two different average stresses. Curves of (0°,±20°n)s laminates intersect at 56
MPa and those of (0°,±30°n)s laminates intersect at 42 MPa, i.e. 49 MPa (average value).
This difference in stress level can be explained by the fact that this second image does not
include the friction effects which are particularly high in (0°,±30°n)s laminates as shown in
the first image.

(a) (c)
(b)

Figure 5. Average stresses in function of the distance to the specimen free edge: (a) normal stress without
friction effects (k=0), (b) shear stress without friction effects (k=0); (c) shear stress with friction effects (k=0.75)

In order to take into account friction effects, we select the value of k to 0.75 according to the
Bassery’s work [6] which investigate on a five harness carbon/PPS satin weave. If we plot the
average shear stress adding to the friction effects as a function of free edge distance, we
obtain the last picture (c). We see that all curves now intersect at the same point which
coordinates are around 0.3mm and 55 MPa. Table 3 summarizes all identified parameters.

y0 [µm] σ zzD [MPa] σ xzD = σ yzD [MPa] k

300(~ply thickness) 25,78 49/55 0,75
Table 1. Summary of criterion parameters; in red value with friction effects.

6
ECCM15 - 15TH EUROPEAN CONFERENCE ON COMPOSITE MATERIALS, Venice, Italy, 24-28 June 2012

4. Criterion validation
4.1. Application to a laboratory test specimen.
To validate our criterion, we choose, firstly, to apply it on a laboratory tensile specimen. The
material is an 8 layers (+20°,+10°2,-50°2)s laminate. The calculation procedures are the
similar procedure described previously.
Simulation results in figure 6 show that the +20°/-10° interface is the most susceptible to
delamination onset. According to curves of the criterion value as a function of applied stress,
delamination occurs firstly at this interface around 400 MPa. Then, in next step, this
numerical result will be verified with experimental result (waiting to receive specimen from
my industrial partner).

Figure 6: out of plane stresses at free edge in thickness direction and criterion value evolution with applied
tensile stress. Vertical grids on the first picture indicate interface positions.

4.2. Application to a structure.

The second application concerns a simple structure specimen, so-called omega, which mesh is
shown in figure 9. It is 300mm long and 30mm wide and consists of eight layers of a quasi-
isotropic (0°,+90°,±45°)s laminate. The numerical simulation concerns at meso scale (ply
scale) which local coordinate are defined at each element of each layer as shown in figure 7.
To apply our criterion, the mesh refinement is requested. However, it is clear that we cannot
refine the whole of specimen since too heavy computational time for a finite elements
calculation. To resolve that, five parts of specimen were defined as shown in figure 7: (a), (b),
(c), (d) and (e). By numerical simulation, we found that the zone (d) represent as the critical
area under load applied. In this reason, the refinement mesh is required (a size of elements
equal to one sixth of the thickness). Then, this area (d) is joined to the rest of the specimen by
using tie contact conditions which does not interfere significantly the stress field. As previous
simulations, node lines are created at the interfaces between plies and the same criterion
computing procedure is applied.

Figure 7: Mesh of the omega part

7
ECCM15 - 15TH EUROPEAN CONFERENCE ON COMPOSITE MATERIALS, Venice, Italy, 24-28 June 2012

The simulation results as shown in figure 8 illustrate that the open mode of delamination is
the most important. Also, the 0°/90° interface is the most sensitive. The curve of criterion
value as a function of displacement lead us to predict the delamination at 0°/+90° interface for
an applied displacement U2 equal to 4.25mm. In the same manner in above, these results will
be experimentally verified as soon I receive the molding part and this paper will be up date.

0 90 +45 -45 -45 +45 90 0

Figure 8. Out of plane stresses at free edge in thickness direction (left) and criterion evolution with applied
displacement (right). Vertical grids on the first picture indicate interface positions

5. Conclusion
In this study, we have successfully identified a delamination criterion and proposed an
approach for a direct application to industrial parts. The first results are encouraging and open
up prospects for a design closer to the damage mechanisms occurring in the material, which
could significantly reduce the safety factors in particularly high for aerospace applications
(six time of standard deviation). However, in order to obtain high reliability in design, several
parameters will be improved. First, the detect system should be improved in order to be more
precise on delamination onset and its propagation. It remains difficult to use AE system in
composite woven since the delamination occurs very early. Otherwise, the mesh generation
software methods would be developed result in the simplification for numerical simulation in
order to apply this criterion in structure specimen.

AKNOLEDGEMENTS
We gratefully aknowledge all partners in the project CRISTAL (Carbone foRge Improved
procesS forTechnological Advance Level) for support to ARMINES Centre des Matériaux.

REFERENCES
[1] Herakovich C. T. On the relationship between engineering properties and delamination of
composite materials. J. Composite materials, vol. 15, pp. 336-348, 1981.
[2] Kim R. Y., Soni S. R. Experimental and analytical studies on the onset of delamination in
laminated composites. Criterion for initiation of delamination. J. Composite materials, vol. 18, pp.
70-80, 1984.
[3] Brewer J. C., Lagace P. A. Quadratic stress criterion for initiation of delamination. J. Composite
materials, vol. 22, pp. 1141-55, 1988.
[4] Lecuyer F., Engrand D. A methodology for the identification of a criterion for delamination
initiation. “Proceeding of 8th National Days of Composites (JNC 8)” Paris, France, 1992.
[5] Pongsak N. Comportement de composites tissés épais soumis à des sollicitations quasi statiques.
Revue des composite et des matériaux avancés, vol. 13, pp. 41-80, 2010
[6] Bassery J., Renard J. Etude du délaminage en fatigue des composites tissés carbone/PPS. Revue
des Composites et des Matériaux Avancés, vol. 10, pp. 345 – 380, 2011.
[7] Bathias Claude. “Une revue des méthodes de caractérisation du délaminage des matériaux
composites.” Annales des Composites, vol. 1, pp. 5-25, 1995.
[8] Lorriot Th. Onset of free-edge delamination in composite laminates under tensile loading.
Composites part B, vol. 34, pp. 459-471, 2003.
[9] Pongsak N., Renard J. Use of acoustic emission to discriminate damage modes in carbon fibre
reinforced epoxy laminate during tensile and buckling loading. “Proceeding of 15th European
Conference on Composite Materials”, Venice, Italy, 24 -28 June 2012.
8